mode='min',
baseline=None) # 早停参数
params = dict(cv=5,
epochs=500,
batch_size=32,
kfold_index=kfold_index,
cb=[cb])
config.extent(params)
channels_names = ['scores7', 'conservation', 'sequence_feature', 'splicing']
(x_train,
y_train), (x_test,
y_test), (x_test_1,
y_test_1), (x_test_2,
y_test_2) = data.get_channels(channels_names)
# 模型搭建
model_file_cv = [
r'./models/scores7/cv{}.h5', r'./models/conservation/cv{}.h5',
r'./models/sequence_feature/cv{}.h5', r'./models/splicing/cv{}.h5'
]
config1 = {
"lr": 1e-04,
"ut_1": 1024,
"l1": 0.0,
"ut_2": 256,
"l2": 0.00,
"dp": 0.0,
'a': 'leaky_relu',
import matplotlib.pyplot as plt
from core.utils import get_kfold_index
from core.data import Data
from core.app_config import AppConfig
from core.scoring import scob
matplotlib.use('Agg')
param = {
"lr": 1e-04,
"ut_1": 1024,
"l1": 0.0,
"ut_2": 256,
"l2": 0.00,
"dp": 0.0,
'a': 'leaky_relu',
'inputs_shape': (7, )
}
model = build_sub_model_1(**param)
data = Data()
(x_train, y_train), (x_test,
y_test), (x_test_1,
y_test_1), (x_test_2,
y_test_2) = data.get_channels(
['sequence_feature'])
model.fit(x_train[0], y_train, validation_split=0.2, batch_size=32, epochs=50)
score = scob.get_scores(y_test_2[:, 1], model.predict(x_test_2)[:, 1])
print(score)
